I've noticed that database performance tuning can become an ongoing cycle. You add indexes, rewrite queries, introduce caching, partition tables, and sometimes denormalize data to squeeze out more performance.

But every optimization also adds something to maintain. More indexes can slow down writes, caching introduces invalidation problems, and denormalization can make data consistency harder to manage.

At some point, I wonder whether the improvement is actually worth the extra complexity.

How do you decide when to stop optimizing a database? Do you have specific performance targets or benchmarks, or does it mostly come down to the workload and the problems users are experiencing?

I'd be interested in hearing about optimizations that genuinely paid off, as well as ones you later regretted.